Markelle Fultz Discussion

[spikeslovechild]

RPM is definitely the closest thing that we have right now to a good all encompassing statistic.

The fact that they keep the formula private is annoying, but the general consensus is that it is pretty similar to RAPM.

Win Shares, BPM ect are generally considered inferior.

I still think it overrates guys on good teams though.

Look I don't know the details of RPM but I do remember getting into similar arguments regarding baseball WAR more then a decade ago. I don't know if the formula is still done the same but it used to be that fangraphs WAR used xFIP while Bref focused on the pitchers ERA and for the hitters they used BABIP the problem is people where using the players WAR figure as evidence they should win awards.

So the questions sort of becomes should you give credit to the player for what did happen versus what should have happened. I always erred on what actually happened and honestly I don't even know how good of predictive formula RPM is there seems to be a fairly large variance of his RPM numbers that isn't caught up in the other stats and honestly looking at his raw numbers there doesn't seem a ton of improvement from 2015 to 2016. His shooting numbers improved slightly to .486 TS%. His AST% improved but his turnover rate spiked at pretty much the same rate which should be a negative.

His OWS stayed exactly the same and his OBPM was actually worse. So maybe one of you could explain how Smart improved his offense last year not slightly but from a net negative to a net positive because it makes zero sense.

Just because other stats are more consistent doesn't make them better. You ask any NBA front office which publicly available stat is the closest thing to the proprietary models that they use to evaluate guys, most of them will say RPM, a few may say BPM and no one will say win shares. It doesn't make it perfect or something you don't have to put into context.

RPM is a relatively complicated multifactor regression. You can't just say "oh it's because of this stat or that stat". It could be that the other point guards around him played worse (Reggie Jackson is a name that immediately jumps out to me). It could be because it values the increase in TS% and Assist rate more than you do. It could be because his teammates randomly played better with him on the court this year. It could be because Stevens had him stand in a different place on offense where teams had to pay more attention to him, opening the floor for other guys. It could be because his usage rate spiked so he gets more "credit" for being part of a top 10 offense. It could be a whole host of reasons. Reverse engineering what the exact reason sort of defeats the purpose of having the statistic.

You are way too hung up on the raw number for his RPM, whether its negative or positive ect. It's a relative number. It doesn't matter whether his number is -20 or 0 or 10,000, if there are 30 people ahead of him in ORPM then there are 30 people ahead of him. And whether he was 30th in ORPM among PGs the way he was this year, or 46th the way he was the year before, either way he is bad on that end, especially for a guy on a good team. But probably he's not quite as destructive to a team's offense as you may think, which makes sense given that the team's production didn't completely fall off a cliff with him in the game.

As far as "predictive value" goes, I can't sit here and tell you "Oh I expect Smart to have a -0.21 RPM this year". But I'm sure there is a certain range that someone with more time/interest could predict with 95% confidence, and I will say I am guessing he will somewhere between the 20th and 40th point guard.

Definitely ahead of Randle, to bring it full circle.

Well to bring it full circle I offered to make a sig bet so if this were true you'd think it would have been taken up on. I guess the problem I have is you guys seem to putting your entire faith in a statistic that is undocumented and that you can't explain nor seem interested in explaining and I'm sorry if every other indicator says Smart is a terrible offensive player and not a good player overall then you need to have some sort of argument other then RPM of why or at the very least be able to explain why RPM has him valued differently.

Saying well thats what RPM says and thats that isn't terribly convincing and I'm open to be convinced. I change my opinions often Okafor is a perfect example. I was high on him due to the way he ended his rookie year prior to getting injured. He was efficient. He had his J working. His FT percentage was improving. I liked his raw talent going forward. Then last year happened and my entire opinion of him changed.

I'm not someone whose afraid to admit his mistakes and change his opinions but there has to be a reason for doing so. So if you guys want to believe Smart is a good player go ahead but I'll wait until he shows it on the court we are what now on the 4th year.

[Sixerscan]

Look I don't know the details of RPM but I do remember getting into similar arguments regarding baseball WAR more then a decade ago. I don't know if the formula is still done the same but it used to be that fangraphs WAR used xFIP while Bref focused on the pitchers ERA and for the hitters they used BABIP the problem is people where using the players WAR figure as evidence they should win awards.

So the questions sort of becomes should you give credit to the player for what did happen versus what should have happened. I always erred on what actually happened and honestly I don't even know how good of predictive formula RPM is there seems to be a fairly large variance of his RPM numbers that isn't caught up in the other stats and honestly looking at his raw numbers there doesn't seem a ton of improvement from 2015 to 2016. His shooting numbers improved slightly to .486 TS%. His AST% improved but his turnover rate spiked at pretty much the same rate which should be a negative.

His OWS stayed exactly the same and his OBPM was actually worse. So maybe one of you could explain how Smart improved his offense last year not slightly but from a net negative to a net positive because it makes zero sense.

Just because other stats are more consistent doesn't make them better. You ask any NBA front office which publicly available stat is the closest thing to the proprietary models that they use to evaluate guys, most of them will say RPM, a few may say BPM and no one will say win shares. It doesn't make it perfect or something you don't have to put into context.

RPM is a relatively complicated multifactor regression. You can't just say "oh it's because of this stat or that stat". It could be that the other point guards around him played worse (Reggie Jackson is a name that immediately jumps out to me). It could be because it values the increase in TS% and Assist rate more than you do. It could be because his teammates randomly played better with him on the court this year. It could be because Stevens had him stand in a different place on offense where teams had to pay more attention to him, opening the floor for other guys. It could be because his usage rate spiked so he gets more "credit" for being part of a top 10 offense. It could be a whole host of reasons. Reverse engineering what the exact reason sort of defeats the purpose of having the statistic.

You are way too hung up on the raw number for his RPM, whether its negative or positive ect. It's a relative number. It doesn't matter whether his number is -20 or 0 or 10,000, if there are 30 people ahead of him in ORPM then there are 30 people ahead of him. And whether he was 30th in ORPM among PGs the way he was this year, or 46th the way he was the year before, either way he is bad on that end, especially for a guy on a good team. But probably he's not quite as destructive to a team's offense as you may think, which makes sense given that the team's production didn't completely fall off a cliff with him in the game.

As far as "predictive value" goes, I can't sit here and tell you "Oh I expect Smart to have a -0.21 RPM this year". But I'm sure there is a certain range that someone with more time/interest could predict with 95% confidence, and I will say I am guessing he will somewhere between the 20th and 40th point guard.

Definitely ahead of Randle, to bring it full circle.

Well to bring it full circle I offered to make a sig bet so if this were true you'd think it would have been taken up on. I guess the problem I have is you guys seem to putting your entire faith in a statistic that is undocumented and that you can't explain nor seem interested in explaining and I'm sorry if every other indicator says Smart is a terrible offensive player and not a good player overall then you need to have some sort of argument other then RPM of why or at the very least be able to explain why RPM has him valued differently.

Saying well thats what RPM says and thats that isn't terribly convincing and I'm open to be convinced. I change my opinions often Okafor is a perfect example. I was high on him due to the way he ended his rookie year prior to getting injured. He was efficient. He had his J working. His FT percentage was improving. I liked his raw talent going forward. Then last year happened and my entire opinion of him changed.

I'm not someone whose afraid to admit his mistakes and change his opinions but there has to be a reason for doing so. So if you guys want to believe Smart is a good player go ahead but I'll wait until he shows it on the court we are what now on the 4th year.

I didn't put my entire faith in a statistic. As I said in that post you quoted/bolded, "It doesn't make it perfect or something you don't have to put into context."

I wasn't trying to convince you or teach you about the statistic, beyond letting you know that the general consensus is that RPM is the best measure that we have right now. Yes the actual formula is murky, which is annoying, but as I said it's modeled off of RAPM and there's ample discussion about what goes into that if you google it. Here's one example.

But since you asked, the issue with win shares and box score plus minus (and PER) is that they are largely (in PER's case entirely) based on box score stats. Which was the best we could do before we had good play by play/5 man lineup data (And it's still the best we can do when comparing to past eras which didn't have that data, which is part of the reason why basketball-reference uses it). But now that we have that, it makes sense to deemphasize box score stats in favor of emphasizing whether or not the team actually scores more/defends better when you are on the court. Because the box score stat-base stats require more assumptions on the part of the person putting the formula together, while play-by-play regression metrics (like RPM) are inherently more data driven. At the end of the day, we don't really care about how many points you score, or even how efficiently you do it, we care whether you make your team better.

Separately, I also disagree that OWS and OBPM have that much different opinions on Smart. According to BBR he was 85th in OWS and and 80th OBPM among "guards" while he was 54th in ORPM among point guards and shooting guards according to ESPN. So those two says he grades out as a bad 3rd guard offensively while ORPM says he's a bad 2nd guard. None of them say that he's "good" imo.

If I had to guess why ORPM is slightly more positive about him, it's that despite his historically awful shooting, the Celtics were a very good offensive team and they didn't do that much worse with Smart on the court. If he was part of a bad offense, or if the Celtics completely submarined whenever he was out there, I would imagine he would grade out much worse.

[Ericb5]

One issue with comparing Smart to other point guards is that not all point guards have the same skills or roles. For example, Sergio Rodriguez and Marcus Smart are so different that comparing them both to point guards is silly.

So when people say that he was Nth in any stat among all point guards, it is inherently flawed. That is just one of the reasons that I don't like statistical arguments about players.

If I had a stat that I called Eric's metric and I just assigned a decimal value to every player in the league based on my opinions, it would be considered a meaningful value if I was right most of the time, and meaningless if I wasn't.

That's basically how it works with these stats. They only are meaningful if they comport with your opinions on the players that are scored.

You guys go back and forth with all of these advanced stats and all that we are left with is that you disagree with each other as to who is better. You might as well just talk about basketball instead of the statistics. You would get to the same place.



Sent from my iPhone using Tapatalk

[Sixerscan]

The lack of being able to fully account for a guy's role is absolutely the biggest issue with advanced statistics in basketball.

I'm just saying, if you are picking and choosing between advanced stats, the stats that use play by play data are better than the ones that rely heavily on the box score. If someone can use sportsvu data or other stuff to better account for roles, that will be even better.

Either way, saying a guy is 30th in RPM isn't supposed to literally mean he's the 30th best, as I have said several times.

[Chris76]

Look I don't know the details of RPM but I do remember getting into similar arguments regarding baseball WAR more then a decade ago. I don't know if the formula is still done the same but it used to be that fangraphs WAR used xFIP while Bref focused on the pitchers ERA and for the hitters they used BABIP the problem is people where using the players WAR figure as evidence they should win awards.

So the questions sort of becomes should you give credit to the player for what did happen versus what should have happened. I always erred on what actually happened and honestly I don't even know how good of predictive formula RPM is there seems to be a fairly large variance of his RPM numbers that isn't caught up in the other stats and honestly looking at his raw numbers there doesn't seem a ton of improvement from 2015 to 2016. His shooting numbers improved slightly to .486 TS%. His AST% improved but his turnover rate spiked at pretty much the same rate which should be a negative.

His OWS stayed exactly the same and his OBPM was actually worse. So maybe one of you could explain how Smart improved his offense last year not slightly but from a net negative to a net positive because it makes zero sense.

Just because other stats are more consistent doesn't make them better. You ask any NBA front office which publicly available stat is the closest thing to the proprietary models that they use to evaluate guys, most of them will say RPM, a few may say BPM and no one will say win shares. It doesn't make it perfect or something you don't have to put into context.

RPM is a relatively complicated multifactor regression. You can't just say "oh it's because of this stat or that stat". It could be that the other point guards around him played worse (Reggie Jackson is a name that immediately jumps out to me). It could be because it values the increase in TS% and Assist rate more than you do. It could be because his teammates randomly played better with him on the court this year. It could be because Stevens had him stand in a different place on offense where teams had to pay more attention to him, opening the floor for other guys. It could be because his usage rate spiked so he gets more "credit" for being part of a top 10 offense. It could be a whole host of reasons. Reverse engineering what the exact reason sort of defeats the purpose of having the statistic.

You are way too hung up on the raw number for his RPM, whether its negative or positive ect. It's a relative number. It doesn't matter whether his number is -20 or 0 or 10,000, if there are 30 people ahead of him in ORPM then there are 30 people ahead of him. And whether he was 30th in ORPM among PGs the way he was this year, or 46th the way he was the year before, either way he is bad on that end, especially for a guy on a good team. But probably he's not quite as destructive to a team's offense as you may think, which makes sense given that the team's production didn't completely fall off a cliff with him in the game.

As far as "predictive value" goes, I can't sit here and tell you "Oh I expect Smart to have a -0.21 RPM this year". But I'm sure there is a certain range that someone with more time/interest could predict with 95% confidence, and I will say I am guessing he will somewhere between the 20th and 40th point guard.

Definitely ahead of Randle, to bring it full circle.

Well to bring it full circle I offered to make a sig bet so if this were true you'd think it would have been taken up on. I guess the problem I have is you guys seem to putting your entire faith in a statistic that is undocumented and that you can't explain nor seem interested in explaining and I'm sorry if every other indicator says Smart is a terrible offensive player and not a good player overall then you need to have some sort of argument other then RPM of why or at the very least be able to explain why RPM has him valued differently.

Saying well thats what RPM says and thats that isn't terribly convincing and I'm open to be convinced. I change my opinions often Okafor is a perfect example. I was high on him due to the way he ended his rookie year prior to getting injured. He was efficient. He had his J working. His FT percentage was improving. I liked his raw talent going forward. Then last year happened and my entire opinion of him changed.

I'm not someone whose afraid to admit his mistakes and change his opinions but there has to be a reason for doing so. So if you guys want to believe Smart is a good player go ahead but I'll wait until he shows it on the court we are what now on the 4th year.

Okafor is a good example, he was doing well his 1st yr. Last year was bad for several reasons.

He plays well this year, would you change your mind again?

He is a good scoring big that should be better at defense.

[Sixerscan]

Okafor is a great example... of how you can have good box score stats and still make your team worse when you are on the court.

[Ericb5]

The lack of being able to fully account for a guy's role is absolutely the biggest issue with advanced statistics in basketball.

I'm just saying, if you are picking and choosing between advanced stats, the stats that use play by play data are better than the ones that rely heavily on the box score. If someone can use sportsvu data or other stuff to better account for roles, that will be even better.

Either way, saying a guy is 30th in RPM isn't supposed to literally mean he's the 30th best, as I have said several times.

If someone says that he is 30th in RPM and someone else is 15th in RPM that doesn't mean that the one guy is better than the other one. It has very little meaning when comparing players to each other.

I agree of course that stats have variable levels of meaningfulness. When you are talking about a rim protecting center for example it becomes pretty clear that team based defensive stats become pretty meaningful.

If you were to use team based defensive stats to compare two point guards though it gets very muddy very quickly.

I think that the best use of advanced stats is to track improvements from within a given team, or within a given player's abilities.

For example, if you are comparing the Sixers when Embiid is at center vs Okafor then you are going to get meaningful stats.

When you want to compare the effectiveness of various lineups then you will get meaningful statistics. Meaning that the team does better or worse in statistical areas when Stauskus plays next to McConnell vs Henderson next to McConnell.

Advanced stats become useful trying to measure improvements or deficits in specific areas for a specific player. Something like Saric shoots X percentage from 3 with a hand in his face one year, and the next year he shoots Y percentage from 3 with a hand in his face.

When deciding what role players to pursue in free agency or who to draft, advanced stats can help to measure where your team weaknesses are. Basically you discover that a given player on your team is more effective when he plays next to a shooter so you pursue someone with that skill.

Or you determine that your center is not an effective rim protector and you think that improving your perimeter defense will decrease the impact of that weakness on your team so you seek out perimeter defenders. In this context someone like Smart becomes much more valuable to your team than a point guard that is good offensively, but poor defensively.




Sent from my iPhone using Tapatalk

[Sixerscan]

The lack of being able to fully account for a guy's role is absolutely the biggest issue with advanced statistics in basketball.

I'm just saying, if you are picking and choosing between advanced stats, the stats that use play by play data are better than the ones that rely heavily on the box score. If someone can use sportsvu data or other stuff to better account for roles, that will be even better.

Either way, saying a guy is 30th in RPM isn't supposed to literally mean he's the 30th best, as I have said several times.

If someone says that he is 30th in RPM and someone else is 15th in RPM that doesn't mean that the one guy is better than the other one. It has very little meaning when comparing players to each other.

I agree of course that stats have variable levels of meaningfulness. When you are talking about a rim protecting center for example it becomes pretty clear that team based defensive stats become pretty meaningful.

If you were to use team based defensive stats to compare two point guards though it gets very muddy very quickly.

I think that the best use of advanced stats is to track improvements from within a given team, or within a given player's abilities.

For example, if you are comparing the Sixers when Embiid is at center vs Okafor then you are going to get meaningful stats.

When you want to compare the effectiveness of various lineups then you will get meaningful statistics. Meaning that the team does better or worse in statistical areas when Stauskus plays next to McConnell vs Henderson next to McConnell.

Advanced stats become useful trying to measure improvements or deficits in specific areas for a specific player. Something like Saric shoots X percentage from 3 with a hand in his face one year, and the next year he shoots Y percentage from 3 with a hand in his face.

When deciding what role players to pursue in free agency or who to draft, advanced stats can help to measure where your team weaknesses are. Basically you discover that a given player on your team is more effective when he plays next to a shooter so you pursue someone with that skill.

Or you determine that your center is not an effective rim protector and you think that improving your perimeter defense will decrease the impact of that weakness on your team so you seek out perimeter defenders. In this context someone like Smart becomes much more valuable to your team than a point guard that is good offensively, but poor defensively.




Sent from my iPhone using Tapatalk

I think if a guy has a clearly better RPM in a fairly similar role it's a relatively strong indicator that he was better that year. More than other stats at least. Of course its not everything, and whether it is predictive for the future is a different question. But saying the fact that a team is better with you on the court is "very little meaning" in determining whether you make your team better is a little much.

I agree those more discrete examples are easier to apply for team building. You shouldn't just take the 12 best RPMs you can find.

[Ericb5]

The lack of being able to fully account for a guy's role is absolutely the biggest issue with advanced statistics in basketball.

I'm just saying, if you are picking and choosing between advanced stats, the stats that use play by play data are better than the ones that rely heavily on the box score. If someone can use sportsvu data or other stuff to better account for roles, that will be even better.

Either way, saying a guy is 30th in RPM isn't supposed to literally mean he's the 30th best, as I have said several times.

If someone says that he is 30th in RPM and someone else is 15th in RPM that doesn't mean that the one guy is better than the other one. It has very little meaning when comparing players to each other.

I agree of course that stats have variable levels of meaningfulness. When you are talking about a rim protecting center for example it becomes pretty clear that team based defensive stats become pretty meaningful.

If you were to use team based defensive stats to compare two point guards though it gets very muddy very quickly.

I think that the best use of advanced stats is to track improvements from within a given team, or within a given player's abilities.

For example, if you are comparing the Sixers when Embiid is at center vs Okafor then you are going to get meaningful stats.

When you want to compare the effectiveness of various lineups then you will get meaningful statistics. Meaning that the team does better or worse in statistical areas when Stauskus plays next to McConnell vs Henderson next to McConnell.

Advanced stats become useful trying to measure improvements or deficits in specific areas for a specific player. Something like Saric shoots X percentage from 3 with a hand in his face one year, and the next year he shoots Y percentage from 3 with a hand in his face.

When deciding what role players to pursue in free agency or who to draft, advanced stats can help to measure where your team weaknesses are. Basically you discover that a given player on your team is more effective when he plays next to a shooter so you pursue someone with that skill.

Or you determine that your center is not an effective rim protector and you think that improving your perimeter defense will decrease the impact of that weakness on your team so you seek out perimeter defenders. In this context someone like Smart becomes much more valuable to your team than a point guard that is good offensively, but poor defensively.




Sent from my iPhone using Tapatalk

I think if a guy has a clearly better RPM in a fairly similar role it's a relatively strong indicator that he was better that year. More than other stats at least. Of course its not everything, and whether it is predictive for the future is a different question. But saying the fact that a team is better with you on the court is "very little meaning" in determining whether you make your team better is a little much.

I agree those more discrete examples are easier to apply for team building. You shouldn't just take the 12 best RPMs you can find.

If a team is better when you are on the floor it has great meaning to the team. I just don't think it has meaning when comparing that person to another person on another team.

If Smart makes the Celtics better when he is on the floor it has no meaning when comparing him to Randle by saying that Randle improves or doesn't improve the Lakers..

If Smart improves his team, and Randle doesn't, it doesn't make Smart a better player because some teams may be improved by having Randle on them. It's all just stuff to talk about.




Sent from my iPhone using Tapatalk

[TTP]

If someone says that he is 30th in RPM and someone else is 15th in RPM that doesn't mean that the one guy is better than the other one. It has very little meaning when comparing players to each other.

I agree of course that stats have variable levels of meaningfulness. When you are talking about a rim protecting center for example it becomes pretty clear that team based defensive stats become pretty meaningful.

If you were to use team based defensive stats to compare two point guards though it gets very muddy very quickly.

I think that the best use of advanced stats is to track improvements from within a given team, or within a given player's abilities.

For example, if you are comparing the Sixers when Embiid is at center vs Okafor then you are going to get meaningful stats.

When you want to compare the effectiveness of various lineups then you will get meaningful statistics. Meaning that the team does better or worse in statistical areas when Stauskus plays next to McConnell vs Henderson next to McConnell.

Advanced stats become useful trying to measure improvements or deficits in specific areas for a specific player. Something like Saric shoots X percentage from 3 with a hand in his face one year, and the next year he shoots Y percentage from 3 with a hand in his face.

When deciding what role players to pursue in free agency or who to draft, advanced stats can help to measure where your team weaknesses are. Basically you discover that a given player on your team is more effective when he plays next to a shooter so you pursue someone with that skill.

Or you determine that your center is not an effective rim protector and you think that improving your perimeter defense will decrease the impact of that weakness on your team so you seek out perimeter defenders. In this context someone like Smart becomes much more valuable to your team than a point guard that is good offensively, but poor defensively.




Sent from my iPhone using Tapatalk

I think if a guy has a clearly better RPM in a fairly similar role it's a relatively strong indicator that he was better that year. More than other stats at least. Of course its not everything, and whether it is predictive for the future is a different question. But saying the fact that a team is better with you on the court is "very little meaning" in determining whether you make your team better is a little much.

I agree those more discrete examples are easier to apply for team building. You shouldn't just take the 12 best RPMs you can find.

If a team is better when you are on the floor it has great meaning to the team. I just don't think it has meaning when comparing that person to another person on another team.

If Smart makes the Celtics better when he is on the floor it has no meaning when comparing him to Randle by saying that Randle improves or doesn't improve the Lakers..

If Smart improves his team, and Randle doesn't, it doesn't make Smart a better player because some teams may be improved by having Randle on them. It's all just stuff to talk about.




Sent from my iPhone using Tapatalk

Definitely disagree here. It shows that Smart is capable of having a positive role within a team and Randle thus far hasn't shown that. That's pretty important when investigating player value.

[Ericb5]

I think if a guy has a clearly better RPM in a fairly similar role it's a relatively strong indicator that he was better that year. More than other stats at least. Of course its not everything, and whether it is predictive for the future is a different question. But saying the fact that a team is better with you on the court is "very little meaning" in determining whether you make your team better is a little much.

I agree those more discrete examples are easier to apply for team building. You shouldn't just take the 12 best RPMs you can find.

If a team is better when you are on the floor it has great meaning to the team. I just don't think it has meaning when comparing that person to another person on another team.

If Smart makes the Celtics better when he is on the floor it has no meaning when comparing him to Randle by saying that Randle improves or doesn't improve the Lakers..

If Smart improves his team, and Randle doesn't, it doesn't make Smart a better player because some teams may be improved by having Randle on them. It's all just stuff to talk about.




Sent from my iPhone using Tapatalk

Definitely disagree here. It shows that Smart is capable of having a positive role within a team and Randle thus far hasn't shown that. That's pretty important when investigating player value.

It's a feather in his cap that he improves his team, but if Randle doesn't improve his team the calculus can't be so crude that it means that Smart is a positive player and Randle is a negative one.

Smart may not improve the Lakers, and Randle may improve the Celtics. I say this of course just as a matter of argument, and not that I think that is necessarily true in the case of those two guys and teams.

People all the time will cite a single statistic to argue that it means that a player is a positive or negative, as if it is a zero sum game. As if a positive player would be a positive on all teams in all contexts, and a negative player would be a negative player on all teams and in all contexts.

I think that the argument should be about basketball, and then statistics can be used contextually to make specific points about a player, but if the whole argument is based on statistics then it is meaningless.

Plus my real beef is to use statistics to compare players to one another. The true value of statistics is to measure a player's effectiveness in the limited context in which he plays. If you are going to compare him to a player on another team then they need to be playing the same role on the same kind of team to derive any meaning from it.





Sent from my iPhone using Tapatalk

[Sixerscan]

If someone says that he is 30th in RPM and someone else is 15th in RPM that doesn't mean that the one guy is better than the other one. It has very little meaning when comparing players to each other.

I agree of course that stats have variable levels of meaningfulness. When you are talking about a rim protecting center for example it becomes pretty clear that team based defensive stats become pretty meaningful.

If you were to use team based defensive stats to compare two point guards though it gets very muddy very quickly.

I think that the best use of advanced stats is to track improvements from within a given team, or within a given player's abilities.

For example, if you are comparing the Sixers when Embiid is at center vs Okafor then you are going to get meaningful stats.

When you want to compare the effectiveness of various lineups then you will get meaningful statistics. Meaning that the team does better or worse in statistical areas when Stauskus plays next to McConnell vs Henderson next to McConnell.

Advanced stats become useful trying to measure improvements or deficits in specific areas for a specific player. Something like Saric shoots X percentage from 3 with a hand in his face one year, and the next year he shoots Y percentage from 3 with a hand in his face.

When deciding what role players to pursue in free agency or who to draft, advanced stats can help to measure where your team weaknesses are. Basically you discover that a given player on your team is more effective when he plays next to a shooter so you pursue someone with that skill.

Or you determine that your center is not an effective rim protector and you think that improving your perimeter defense will decrease the impact of that weakness on your team so you seek out perimeter defenders. In this context someone like Smart becomes much more valuable to your team than a point guard that is good offensively, but poor defensively.




Sent from my iPhone using Tapatalk

I think if a guy has a clearly better RPM in a fairly similar role it's a relatively strong indicator that he was better that year. More than other stats at least. Of course its not everything, and whether it is predictive for the future is a different question. But saying the fact that a team is better with you on the court is "very little meaning" in determining whether you make your team better is a little much.

I agree those more discrete examples are easier to apply for team building. You shouldn't just take the 12 best RPMs you can find.

If a team is better when you are on the floor it has great meaning to the team. I just don't think it has meaning when comparing that person to another person on another team.

If Smart makes the Celtics better when he is on the floor it has no meaning when comparing him to Randle by saying that Randle improves or doesn't improve the Lakers..

If Smart improves his team, and Randle doesn't, it doesn't make Smart a better player because some teams may be improved by having Randle on them. It's all just stuff to talk about.




Sent from my iPhone using Tapatalk

No offense but what are you basing this on? Just that you have decided that this is how it works? Because you are being very dogmatic about this...

I'm not saying the two situations are identical, and I agree that it's less valuable the less similar the players' roles are*, but it's not like we are attempting to predict weather patterns in the atmospheres of two different planets 100 million miles apart. We're comparing 5 people playing basketball against 5 other people playing basketball. There's a decent amount of similarity and carry over.

Yes there are variables but that's literally what regression analysis is for...

*Which is why when comparing a point guard like Smart to a power forward like Randle I care much less about the raw number than how they compare to other people at their positions.

[spikeslovechild]

If a team is better when you are on the floor it has great meaning to the team. I just don't think it has meaning when comparing that person to another person on another team.

If Smart makes the Celtics better when he is on the floor it has no meaning when comparing him to Randle by saying that Randle improves or doesn't improve the Lakers..

If Smart improves his team, and Randle doesn't, it doesn't make Smart a better player because some teams may be improved by having Randle on them. It's all just stuff to talk about.




Sent from my iPhone using Tapatalk

Definitely disagree here. It shows that Smart is capable of having a positive role within a team and Randle thus far hasn't shown that. That's pretty important when investigating player value.

It's a feather in his cap that he improves his team, but if Randle doesn't improve his team the calculus can't be so crude that it means that Smart is a positive player and Randle is a negative one.

Smart may not improve the Lakers, and Randle may improve the Celtics. I say this of course just as a matter of argument, and not that I think that is necessarily true in the case of those two guys and teams.

People all the time will cite a single statistic to argue that it means that a player is a positive or negative, as if it is a zero sum game. As if a positive player would be a positive on all teams in all contexts, and a negative player would be a negative player on all teams and in all contexts.

I think that the argument should be about basketball, and then statistics can be used contextually to make specific points about a player, but if the whole argument is based on statistics then it is meaningless.

Plus my real beef is to use statistics to compare players to one another. The true value of statistics is to measure a player's effectiveness in the limited context in which he plays. If you are going to compare him to a player on another team then they need to be playing the same role on the same kind of team to derive any meaning from it.

Sent from my iPhone using Tapatalk

This I agree with which is why I try to look at both boxscore and stats like WS/48 and BPM to get a complete picture. There are some situations that are bad fits where the player plays well but the team doesn't. For example Zach Randolph on the knicks he was playing with Eddy Curry it was just a bad pairing not only from an offensive standpoint but a defensive standpoint as well.

Almost all his numbers shot up way up after joining the Grizz.

On the flip side you have guys like Chandler and Deandre Jordan two centers who really thrived having a PG who got them the ball. They wouldn't be as nearly as valuable playing for another team.

[Ericb5]

I think if a guy has a clearly better RPM in a fairly similar role it's a relatively strong indicator that he was better that year. More than other stats at least. Of course its not everything, and whether it is predictive for the future is a different question. But saying the fact that a team is better with you on the court is "very little meaning" in determining whether you make your team better is a little much.

I agree those more discrete examples are easier to apply for team building. You shouldn't just take the 12 best RPMs you can find.

If a team is better when you are on the floor it has great meaning to the team. I just don't think it has meaning when comparing that person to another person on another team.

If Smart makes the Celtics better when he is on the floor it has no meaning when comparing him to Randle by saying that Randle improves or doesn't improve the Lakers..

If Smart improves his team, and Randle doesn't, it doesn't make Smart a better player because some teams may be improved by having Randle on them. It's all just stuff to talk about.




Sent from my iPhone using Tapatalk

No offense but what are you basing this on? Just that you have decided that this is how it works? Because you are being very dogmatic about this...

I'm not saying the two situations are identical, and I agree that it's less valuable the less similar the players' roles are*, but it's not like we are attempting to predict weather patterns in the atmospheres of two different planets 100 million miles apart. We're comparing 5 people playing basketball against 5 other people playing basketball. There's a decent amount of similarity and carry over.

Yes there are variables but that's literally what regression analysis is for...

*Which is why when comparing a point guard like Smart to a power forward like Randle I care much less about the raw number than how they compare to other people at their positions.

What I'm basing it on is just my opinion of why I distrust and dislike statistical arguments. I realize that analytics is an attempt to put evidence behind opinions, but I think that the evidence leads to incorrect assumptions enough of the time to make it basically just another form of opinion that is right or wrong based on the outcomes.

People think that they have a statistical representation of whether someone is positive or negative, and then they use that number to say that some other player in a different context is better or worse, and I think that the significance of the statistics is way over stated. If both players have the same skill sets, roles, and contexts then the stats are more relevant.

When I make this point about a big man having a similar defensive rating as a point guard and how using their argument that should mean that they are equal defenders, people will quickly revert to an idea that the big man is SUPPOSED to be a defender and the point guard isn't so the same defensive rating actually means that the big man is worse. They acknowledge the contextual differences while still using the stat as evidence anyway.

Just because the goal of regression analysis is to correct for the missing context doesn't mean that it accomplishes that.

I work in IT and I see it all the time where someone will find an application that works for their needs, and then they start to try to use the same application for other needs as if the utility of the first use case will logically extend to other use cases. That's the fallacy that I'm highlighting. When you are holding a hammer, every problem starts to look like a nail.

Just because statistics provide value in certain contexts doesn't mean that you can derive the same utility from all contexts. When you get right down to it, you are just stating your own opinions, and that is what all of us are doing.

Too many people fall in love with statistics and use them outside of their context as evidence of something which really isn't evidence. It is just opinion. What I base my opinions on is the same thing that you base your opinions on which is decades of watching basketball passionately.

I'm all for collecting the data. I just don't want to extrapolate too far with the data.

Sent from my iPhone using Tapatalk

[Sixerscan]

I've said like 50 times that you have to account for contexts?

I'm all for not assuming everything is a nail. It just seems like you have made the blanket assumption that this can't be a nail without looking into the viability of the hammer at all. Which is fine I guess, it just sort of seems to me that you aren't that much different than those hammer wielders you seem to dislike.

[spikeslovechild]

If a team is better when you are on the floor it has great meaning to the team. I just don't think it has meaning when comparing that person to another person on another team.

If Smart makes the Celtics better when he is on the floor it has no meaning when comparing him to Randle by saying that Randle improves or doesn't improve the Lakers..

If Smart improves his team, and Randle doesn't, it doesn't make Smart a better player because some teams may be improved by having Randle on them. It's all just stuff to talk about.




Sent from my iPhone using Tapatalk

No offense but what are you basing this on? Just that you have decided that this is how it works? Because you are being very dogmatic about this...

I'm not saying the two situations are identical, and I agree that it's less valuable the less similar the players' roles are*, but it's not like we are attempting to predict weather patterns in the atmospheres of two different planets 100 million miles apart. We're comparing 5 people playing basketball against 5 other people playing basketball. There's a decent amount of similarity and carry over.

Yes there are variables but that's literally what regression analysis is for...

*Which is why when comparing a point guard like Smart to a power forward like Randle I care much less about the raw number than how they compare to other people at their positions.

What I'm basing it on is just my opinion of why I distrust and dislike statistical arguments. I realize that analytics is an attempt to put evidence behind opinions, but I think that the evidence leads to incorrect assumptions enough of the time to make it basically just another form of opinion that is right or wrong based on the outcomes.

People think that they have a statistical representation of whether someone is positive or negative, and then they use that number to say that some other player in a different context is better or worse, and I think that the significance of the statistics is way over stated. If both players have the same skill sets, roles, and contexts then the stats are more relevant.

When I make this point about a big man having a similar defensive rating as a point guard and how using their argument that should mean that they are equal defenders, people will quickly revert to an idea that the big man is SUPPOSED to be a defender and the point guard isn't so the same defensive rating actually means that the big man is worse. They acknowledge the contextual differences while still using the stat as evidence anyway.

Just because the goal of regression analysis is to correct for the missing context doesn't mean that it accomplishes that.

I work in IT and I see it all the time where someone will find an application that works for their needs, and then they start to try to use the same application for other needs as if the utility of the first use case will logically extend to other use cases. That's the fallacy that I'm highlighting. When you are holding a hammer, every problem starts to look like a nail.

Just because statistics provide value in certain contexts doesn't mean that you can derive the same utility from all contexts. When you get right down to it, you are just stating your own opinions, and that is what all of us are doing.

Too many people fall in love with statistics and use them outside of their context as evidence of something which really isn't evidence. It is just opinion. What I base my opinions on is the same thing that you base your opinions on which is decades of watching basketball passionately.

I'm all for collecting the data. I just don't want to extrapolate too far with the data.

Sent from my iPhone using Tapatalk

I think you are sort of misrepresenting the reasons why people fall in love with certain applications. They do so in most cases because they know how to use it or in your case know how to build a solution. They don't want to be challenged to fill in those gaps of knowledge or reading whitepapers because that takes time and takes them out of their comfort level.

We should never be too comfortable with our own opinions where we should be afraid to have our belief system challenged. Don't be one of those people still running beehive as your collaboration platform because thats what you know. If that sounds strangely specific it is.

Stats are useful for settling arguments because in the case of a stalemate it sort of ends well I believe this, well thats not true I believe that and puts the onus on someone to explain their position.

[spikeslovechild]

Just because other stats are more consistent doesn't make them better. You ask any NBA front office which publicly available stat is the closest thing to the proprietary models that they use to evaluate guys, most of them will say RPM, a few may say BPM and no one will say win shares. It doesn't make it perfect or something you don't have to put into context.

RPM is a relatively complicated multifactor regression. You can't just say "oh it's because of this stat or that stat". It could be that the other point guards around him played worse (Reggie Jackson is a name that immediately jumps out to me). It could be because it values the increase in TS% and Assist rate more than you do. It could be because his teammates randomly played better with him on the court this year. It could be because Stevens had him stand in a different place on offense where teams had to pay more attention to him, opening the floor for other guys. It could be because his usage rate spiked so he gets more "credit" for being part of a top 10 offense. It could be a whole host of reasons. Reverse engineering what the exact reason sort of defeats the purpose of having the statistic.

You are way too hung up on the raw number for his RPM, whether its negative or positive ect. It's a relative number. It doesn't matter whether his number is -20 or 0 or 10,000, if there are 30 people ahead of him in ORPM then there are 30 people ahead of him. And whether he was 30th in ORPM among PGs the way he was this year, or 46th the way he was the year before, either way he is bad on that end, especially for a guy on a good team. But probably he's not quite as destructive to a team's offense as you may think, which makes sense given that the team's production didn't completely fall off a cliff with him in the game.

As far as "predictive value" goes, I can't sit here and tell you "Oh I expect Smart to have a -0.21 RPM this year". But I'm sure there is a certain range that someone with more time/interest could predict with 95% confidence, and I will say I am guessing he will somewhere between the 20th and 40th point guard.

Definitely ahead of Randle, to bring it full circle.

Well to bring it full circle I offered to make a sig bet so if this were true you'd think it would have been taken up on. I guess the problem I have is you guys seem to putting your entire faith in a statistic that is undocumented and that you can't explain nor seem interested in explaining and I'm sorry if every other indicator says Smart is a terrible offensive player and not a good player overall then you need to have some sort of argument other then RPM of why or at the very least be able to explain why RPM has him valued differently.

Saying well thats what RPM says and thats that isn't terribly convincing and I'm open to be convinced. I change my opinions often Okafor is a perfect example. I was high on him due to the way he ended his rookie year prior to getting injured. He was efficient. He had his J working. His FT percentage was improving. I liked his raw talent going forward. Then last year happened and my entire opinion of him changed.

I'm not someone whose afraid to admit his mistakes and change his opinions but there has to be a reason for doing so. So if you guys want to believe Smart is a good player go ahead but I'll wait until he shows it on the court we are what now on the 4th year.

Okafor is a good example, he was doing well his 1st yr. Last year was bad for several reasons.

He plays well this year, would you change your mind again?

He is a good scoring big that should be better at defense.

Yes, if he makes the necessary changes my evaluation of him will change. Heck, as will Smarts.

The question becomes sort of how likely it is to happen and with both players I have sort of lost faith in them for different reasons. If you want I can go deeper into why but I've explained it ad naseum already so I prefer if I didn't have to.

[Ericb5]

I've said like 50 times that you have to account for contexts?

I'm all for not assuming everything is a nail. It just seems like you have made the blanket assumption that this can't be a nail without looking into the viability of the hammer at all. Which is fine I guess, it just sort of seems to me that you aren't that much different than those hammer wielders you seem to dislike.

I'm not really specifically talking to/about you. I'm responding to the stat barrage pissing matches that were happening in this thread.

One side throws out a bunch of stats, and the other side responds with a bunch of stats, and they both think that they have evidence on their side, yet the stats are conflicting.

I'm not saying that I am more correct than anyone else. I'm just talking about opinions like everyone else, but some of the stat geeks seem to believe that they aren't talking about opinions, and that they are in fact showing the other side pure EVIDENCE. In reality, the interpretation of the advanced stats are also subject to opinions. The stats don't necessarily show what they purport to show. What they actually show is open to interpretation.

Using someone's shooting percentage is subjective in the same way that saying what your opinion is of his shooting ability. I could say that someone is a good shooter because I watch him, and you could say that he is a good shooter because his shooting percentage is X, and BOTH statements are opinions.

[Ericb5]

No offense but what are you basing this on? Just that you have decided that this is how it works? Because you are being very dogmatic about this...

I'm not saying the two situations are identical, and I agree that it's less valuable the less similar the players' roles are*, but it's not like we are attempting to predict weather patterns in the atmospheres of two different planets 100 million miles apart. We're comparing 5 people playing basketball against 5 other people playing basketball. There's a decent amount of similarity and carry over.

Yes there are variables but that's literally what regression analysis is for...

*Which is why when comparing a point guard like Smart to a power forward like Randle I care much less about the raw number than how they compare to other people at their positions.

What I'm basing it on is just my opinion of why I distrust and dislike statistical arguments. I realize that analytics is an attempt to put evidence behind opinions, but I think that the evidence leads to incorrect assumptions enough of the time to make it basically just another form of opinion that is right or wrong based on the outcomes.

People think that they have a statistical representation of whether someone is positive or negative, and then they use that number to say that some other player in a different context is better or worse, and I think that the significance of the statistics is way over stated. If both players have the same skill sets, roles, and contexts then the stats are more relevant.

When I make this point about a big man having a similar defensive rating as a point guard and how using their argument that should mean that they are equal defenders, people will quickly revert to an idea that the big man is SUPPOSED to be a defender and the point guard isn't so the same defensive rating actually means that the big man is worse. They acknowledge the contextual differences while still using the stat as evidence anyway.

Just because the goal of regression analysis is to correct for the missing context doesn't mean that it accomplishes that.

I work in IT and I see it all the time where someone will find an application that works for their needs, and then they start to try to use the same application for other needs as if the utility of the first use case will logically extend to other use cases. That's the fallacy that I'm highlighting. When you are holding a hammer, every problem starts to look like a nail.

Just because statistics provide value in certain contexts doesn't mean that you can derive the same utility from all contexts. When you get right down to it, you are just stating your own opinions, and that is what all of us are doing.

Too many people fall in love with statistics and use them outside of their context as evidence of something which really isn't evidence. It is just opinion. What I base my opinions on is the same thing that you base your opinions on which is decades of watching basketball passionately.

I'm all for collecting the data. I just don't want to extrapolate too far with the data.

Sent from my iPhone using Tapatalk

I think you are sort of misrepresenting the reasons why people fall in love with certain applications. They do so in most cases because they know how to use it or in your case know how to build a solution. They don't want to be challenged to fill in those gaps of knowledge or reading whitepapers because that takes time and takes them out of their comfort level.

We should never be too comfortable with our own opinions where we should be afraid to have our belief system challenged. Don't be one of those people still running beehive as your collaboration platform because thats what you know. If that sounds strangely specific it is.

Stats are useful for settling arguments because in the case of a stalemate it sort of ends well I believe this, well thats not true I believe that and puts the onus on someone to explain their position.

That's fair. I just want people that form their arguments around stats to accept that they are talking about opinions too.

[LongLiveHinkie]

This thread would make the Night King pack his bags and go home.